Made With Jev

Jev Engineering is a way to build AI agents: an LLM writes, Jev decides, and code acts. Every point where an agent picks a worker, scores a source or approves a tool call goes to Jev, TypeSafe AI’s System One model, instead of to a model that generates text.

The projects on Made with Jev cluster around eight broad areas: agents and browsers, games and real-time applications, triage and routing, trading and markets, content and growth, research and data, robotics and devices, and general tools and apps.

A lot of projects use Jev as the fast decision layer inside browser automation. Examples include flight-search agents, browser agents that choose the next click, LinkedIn recruiting agents, and tools that navigate websites from natural-language instructions. One listed flight-search build reports roughly 7 seconds and $0.004 per run.

Yes. Projects include ad analysis, lead scoring, outreach evaluation, social-post scoring, growth analysis, and an AI-slop detector. One example analyzed 724 ads from 37 brands in about 40 seconds for a reported $0.09 in token costs; another evaluated 3,282 posts for growth patterns.

The directory includes things like Ableton Live voice control, pixel-art generation from Jev probabilities, speech analysis, dynamic forms, desktop automation, website-quality checking, and robot-control experiments. There is even a MuJoCo robotics workbench with Jev in the decision loop.